1
Article Title: The Dynamics of Cell Proliferation
John F. Moxnes
Department for Protection and Materiel
Norwegian Defence Research Establishment
P.O. Box 25
N-2007 Kjeller, Norway
E-mail: [email protected]
Tel.: +47 63 807514, Tel. Dept: +47 63 807000, Fax: +47 63 807509
Johan Haux
Department of Hematology
KSS
SE-541 85 Skövde, Sweden
E-mail: [email protected]
Tel: +46 500 431000
Kjell Hausken
School of Economics, Culture and Social Sciences
University of Stavanger
P.O. Box 2557 Ullandhaug
N-4091 Stavanger, Norway
E-mail: [email protected]
Tel.: +47 51 831632, Tel. Dept: +47 51 831500, Fax: +47 51 831550
Version: December 17, 2003
Keywords: Cells, cancer, Digitoxin, proliferation, differential equations
2
Article Title: The Dynamics of Cell Proliferation
Abstract
The article provides a mathematical description based on the theory of differential equations,
for the proliferation of malignant cells (cancer). A model is developed which enables us to
describe and predict the dynamics of cell proliferation much better than by using ordinary
curve fitting procedures. By using differential equations the ability to foresee the dynamics of
cell proliferation is in general much better than by using polynomial extrapolations. Complex
time relations can be revealed. The mass of each living cell and the number of living cells are
described as functions of time, accounting for each living cell’s age since cell-birth. The
linkage between micro dynamics and the population dynamics is furnished by coupling the
mass increase of each living cell up against the mitosis rate. A comparison is made by in vitro
experiments with cancer cells exposed to digitoxin, a new promising anti cancer drug.
Theoretical results for the total number of cells (living or dead) is found to be in good
agreement with experiments for the cell line considered, assuming different concentrations of
digitoxin. It is shown that for the chosen cell line, the proliferation is halted by an increased
time from birth to mitosis of the cells. The delay is probably connected with changes in the
Ca concentration inside the cell. The enhanced time between the birth and mitosis of a cell
leads effectively to smaller mitosis rates and thereby smaller proliferation rates. This
mechanism is different from the earlier results on digitoxin for different cell lines where an
increased rate of apoptosis was reported. But we find it reasonable that cell lines can react
differently to digitoxin. A development from enhanced time between birth and mitosis to
apoptosis can be furnished, dependent of the sensitivity of the cell lines. This mechanism is
in general very different from the mechanism appealed to by standard chemotherapy and
radiotherapy where the death ratios of the cells are mainly affected. Thus the analysis
supports the view that a quite different mechanism is invoked when using digitoxin. This is
important, since by appealing to different types of mechanism in parallel during cancer
treatment, more selectivity in the targeting of benign versus malignant cells can be invoked.
This increases the probability of successful treatment. The critical digitoxin level
concentration, i.e. the concentration level where the number of living cells is not increasing,
is approximately 50 ng/ml for the cell line we investigated in this article. Therapeutic plasma
concentration of digitoxin when treating cardiac congestion is about 15-33ng/ml, but
individual tolerances are large. The effect of digitoxin during cancer treatment is therefore
3
very promising. The dynamic model constitutes a new powerful tool, supported by empirics,
describing the mechanism or process by which the number of malignant cells during
anticancer treatment can be studied and reduced.
4
1 Introduction
A considerable amount of experiments has been carried out to test the growth and
proliferation of cells. Although this literature is quantitative and technical in nature, with
suggested polynomials describing the cell growth and cell proliferation, lacking is to our
knowledge a mathematical differential equation describing the process. By using differential
equations the ability to foresee the dynamics of cell proliferation is much better than by using
polynomial extrapolations. Complex time relations can be revealed.
Of fundamental importance for all living cells is the ability to divide (mitosis) or to die by
apoptotic or necrotic death. During treatments of living organisms possessing malignant cells
(cancer), chemotherapy and/or radiotherapy are/is frequently used. Cells are very sensitive to
cytostatics or radiation during the G2-M phase in the cell cycle. Therefore the increased
relative amount of deaths caused by the treatment during a short time interval is proportional
with the mitosis rate. The increased death rate caused by chemotherapy and/or radiotherapy
for cell lines with large mitosis rates will then lead to a large negative value of the mitosis
rate minus the death rate, i.e. a strongly negative proliferation rate. Disappointingly, this is
also an obstacle to treatment since benign cells with mitosis rates of the same or higher order
will be strongly attacked and reduced in number and functional quality by such a treatment.
This is due to the non-selective action of most of the standard chemotherapeutic drugs or
radiation. If the mitosis rate is larger than the death rate, and both are small during cancer
development, the corresponding negative proliferation rate caused by treatment is small, and
the number of cells are therefore almost unaffected during “treatment”. By using small
amounts of cytostatics over long time spans, the benign cells with high mitosis/deaths rates
are more affected than the malignant cells with low proliferation.
It is well known that increased Ca ions concentrations in the interior of cells strongly enhance
the probability of so called apoptotic deaths of the cells (Russo 1982). Inhibition of the Na-K
pump can indirectly increase the apoptotic death rates of cells since the reduced effect of the
Na-K pump leads to higher Na concentration in the cells’ interior. Therefore the effect of the
NA-Ca exchanger which pumps Na ions in and Ca ions out of the cell by use of the Na
gradient over the cell membrane, is reduced. This leads to higher Ca concentration in the
cells’ interior and increased probability of apoptotic deaths. Recently, analysis of herbal
extracts used in alternative medicine has revealed that some of these also contain cardiac
5
glycosides, which have been known for a long time to inhibit proliferation of cancer cells by
inhibition of the Na-K pump across the cell membrane. Inspired by the results presented on
the anti-cancer effects of cardiac glycosides, Haux et al. (1999) examined the effect of the
cardiac glycoside digitoxin, and more specifically the effect of digitoxin on different but
typically malignant cell lines. They showed that digitoxin inhibited the proliferation of cells
for most of the malignant cell lines by increasing the number of apoptotic cells. However,
and interestingly, the normal cell lines consisting of lines with large mitosis/deaths rate and
slow mitosis/death rates, were not affected by the digitoxin treatment. This means that a new
sort of selectivity in the targeting of benign versus malignant cells, which is not directly
linked to the proliferation rates of cell lines, is invoked when using digitoxin. In the
experiments the concentration of digitoxin was not higher than in standard treatments of
cardiac diseases. Therapeutic digitoxin concentration does not seem to give any bad side
effects in persons with or without cardiac diseases (Grossmann 1998). Further, Stenkvist et al
(1979) found that five years after mastectomy the recurrences among breast cancer patients
not taking digitalis were ten times that in patients taking digitalis. Also, Moxnes and Hausken
(2003) provided a mathematical dynamic description of the interaction between the organism
and the drug, analysing the dynamics by using ordinary differential equations.
Inspired by these very promising results, this article builds a mathematical model in order to
study the proliferation of cells in a more fundamental way. For the cell line the model’s
intrinsic predictive power is used to analyze the effect of using digitoxin. Of special interest
is the examination of whether the cell proliferation when using digitoxin follows from very
simple analytical relationships where the effect of digitoxin can be captured by a simple rescaling of the death coefficient of the cells. This mechanism is typically invoked during
standard chemotherapy and radiotherapy. The analysis supports the view that a quite different
mechanism is invoked when using digitoxin. This is important, since by appealing to
different types of mechanism in parallel during cancer treatment, more selectivity in the
targeting of benign versus malignant cells can be invoked. This increases the probability of
successful treatment. By using the mathematical model together with the experimental results
we find for our cell line that the proliferation is halted by an increased time from birth to
mitosis of the cells. The enhanced time between the birth and mitosis of a cell leads
effectively to smaller proliferation rates. This mechanism is somewhat different from the
earlier results on digitoxin for different cell lines where an increased rate of apoptosis was
6
reported. But we find it reasonable that cell lines can react differently to digitoxin. A
development from enhanced time between birth and mitosis to apoptosis can be furnished,
dependent of the digitoxin sensitivity of the cell lines.
Further, when introducing digitoxin the simulations are run 30-40 days into the future, and
model is used to pinpoint the exact critical digitoxin level where the number of living cells is
decreasing as a function of time.
Simulations are run over 40 days with varying levels of digitoxin to pinpoint that exact
critical digitoxin level where the number of living cells is decreasing as a function of time.
Section 2 provides the theoretical model. Section 3 presents analytical solutions for simple
scenarios. Section 4 compares the model with results from in vitro experiments with cancer
cells exposed to digitoxin. Section 5 concludes.
2 The theoretical model
Central for the theoretical model are two quantities, the number of cells at different ages, and the
mass of cells of different ages. Let N (t ,τ ) be the number of living cells at time t of ages less than
or equal to τ in a volume V1. Define the age density ρ (t ,τ ) by
def
τ
ρ (t ,τ ) = ∂N (t ,τ ) / ∂τ , i.e. N (t ,τ ) = ∫ ρ (t , u )du,
(2.1)
0
where “def” means a definition. ρ (t ,τ )dτ is the number of living cells with ages in the
interval from τ to τ+dτ. At time t the total number of living cells of all ages is given by
def
N T (t ) = N (t , ∞).
where the subscript “T” is used to indicate the total number of living cells of all ages.
(2.2)
7
Let M (t ,τ ) be the total mass of all living cells at time t of ages less than or equal to τ. Define the
age mass density m(t ,τ ) by
def
τ
m(t ,τ ) = ∂M (t ,τ ) / ∂τ , i.e. M (t ,τ ) = ∫ m(t , u )du
(2.3)
0
where m(t ,τ )dτ is the total mass of all living cells at time t of ages in the interval from τ to
τ+dτ. At time t the total mass of all living cells of all ages, i.e. total living biomass in the
sample volume, is given by
def
M T (t ) = M (t , ∞).
(2.4)
We now define the important ratio
def
m(t , τ )
mc (t , τ ) =
ρ (t ,τ )
def
when ρ (t ,τ ) ≠ 0, mc (t ,τ ) = 0 when ρ (t ,τ ) = 0,
(2.5)
which can be interpreted as the average mass of one living cell at time t of age τ .
The rest of this article focuses on the construction of mathematical models of the two main
quantities; the age density ρ (t ,τ ) and the average mass of a cell mc (t ,τ ) .
For the age density ρ (t ,τ ) in (2.1) the following equation follows directly from the
conservation of cells
The number of living cells at time t+dt of ages between τ+dτ and τ+2dτ equals the number
ρ (t ,τ )dτ of living cells at time t of ages from τ to τ+dτ, plus the number r (t ,τ )dτ dt of living
1
This article applies expected values for all variables and we suppress the word expectation.
8
cells reinforced from t to t+dt of ages between τ and τ+dτ, minus the number l (t ,τ )dτ dt of
living cell lost from t to t+dt of ages between τ and τ+dτ. This gives
ρ (t + dt , τ + dτ )dτ = ρ (t ,τ )dτ + r (t ,τ )dτ dt − l (t ,τ )dτ dt.
(2.6)
Taylor expansion of (2.6) up to order O(h), where time t and age τ is coupled such that dt=dτ,
applying that dt=dτ tends to zero, yields the conservation equation
def
∂ρ (t ,τ ) ∂ρ (t ,τ )
0
+
= r (t ,τ ) − l (t ,τ ), ρ (τ ) = ρ (0,τ )
∂t
∂τ
(2.7)
ρ (0, τ ) is the initial age density. The following differential equation follows directly from mass
conservation of each living cell
def
mc (t + dt ,τ + dτ ) = mc (t ,τ ) + rc (t ,τ )dt − lc (t ,τ )dt ,
mc0 (t )
= mc (t , 0)
(2.8)
(2.8) expresses that the mass of a living cell at time t+dt of ages τ+dτ equals the mass of the
living cell at time t of age τ plus the mass increase rc (t ,τ )dt from t to t+dt, minus the mass loss
lc (t ,τ )dt from t to t+dt. mc (t , 0) is the mass of an average cell at birth. This mass can vary with
time due to different external or internal conditions.
Taylor expansion of (2.8) up to order O(h), where time t and age τ is coupled such that dt=dτ,
applying that dt=dτ tends to zero, yields
def
∂mc (t ,τ ) ∂mc (t ,τ )
+
= rc (t ,τ ) − lc (t ,τ ), M c (t ) =
∂t
∂τ
∞
∫0
mc (t ,τ )dτ ,
where M c (t ) is the total mass of a living cell during the life cycle.
(2.9)
9
Let us now move on to the specific relations that are connected to the life cycle of the cells.
Moving back to equation (2.7), the loss and reinforcement are given as
mod
mod
l (t , τ ) = µ (t ,τ ) ρ (t ,τ ), r (t ,τ ) = 0,
(2.10)
where “ mod ” means that this is a testable model assumption. µ (t ,τ ) is the death rate
coefficient as a function of time (dependent on external conditions) and age2. A living cell’s
death rate usually increases with age τ but also depends on growth mechanisms and whether the
living cell is located in a nutritionally optimal environment at time t. The cell death coefficient
µ (t ,τ ) can generally be divided in two parts, one from apoptotic cell death and one from necrotic
cell death. We do not separate those two events in this article. r(t, τ) is the general
reinforcement. This term will be set to zero in this article since all reinforcements will be
given through a boundary condition for the age density at age zero ρ (t , 0) .
We assume that the loss and reinforcement of mass for an average cell follow the equations
mod
mod
n
rc (t ,τ ) = c1mc (t ,τ ) , lc (t ,τ ) = c2 mc (t ,τ ) − lmit (t ,τ ).
(2.11)
The mass increase mc (t ,τ ) n is due to anabolism, where n is a parameter3, the loss term lc (t ,τ )
is divided in two parts. c2 mc (t ,τ ) is due to catabolism, and lmit (t ,τ ) is due to mitosis loss of a
living cell, where c1 and c2 are parameters which depend on the temperature and the
supply/availability of oxygen and other nutrients.
Crucial for cell division is that a living cell can generally only divide when it has reached a
certain mass size and a specific age. Also the population densities of cells inhibit mitosis due to
contact inhibition. The following mathematical construction is descriptive
2
Strictly speaking, without a more specific relation for this coefficient this equation is a pure definition.
For the distinct process of anabolism there is some discussion in the literature (see e.g. Bartilanffy 1955, 1962)
of whether n=1 or n=2/3. We find empirical support for n=1 which is used in section 3 and thereafter.
3
10
mod
lmit (t ,τ ) = ( rc (t ,τ ) − c2 mc (t ,τ ) ) g (τ )
mod 
 0, when ( rc (t ,τ ) − c2 mc (t ,τ ) ) ≤ 0 or
g (τ ) = 
2 2
 χ (1 − e −τ β ) otherwise, 0 ≤ χ ≤ 1.
(2.12)
NT (t ) ≥ N max
(2.12) states that a living cell divides only if the mass tends to increase, i.e. a part of the mass
increase is converted to offspring. Due to contact inhibition, the cell divides only if the
population density is below a certain value N max . The cell starts to divide at the age of
approximately β . The χ value expresses the mass fraction of the cell growth that is used to
mitosis. χ = 1 means that the cell converts all its potential mass growth into mitosis.
In order to close the system a connection between the mass loss due to mitosis and the number of
offspring must be established. One simple relation is
mod
mcb (t , τ ) = α mc (t ,τ ),
(2.13)
where mcb (t ,τ ) is the mass of a newly born cell at time t, born from parents of age τ . Different
age classes have different average mass for cells, and large cells in general give larger offspring
than smaller cells. The offspring are a constant fraction α of the cells mass.
The number of cells born is now given by the boundary condition
mod ∞
ρ (t , 0) =
∫ ( lmit (t ,τ ) ρ (t ,τ ) / mc (t ,τ ) ) dτ ,
b
ε > 0,
(2.14)
ε
where ε is arbitrarily small but positive (to indicate that the integration does not include zero).
Equation (2.14) expresses that the reinforced (born) number of cells of age zero is equal with the
mass loss lmit (t ,τ ) pr time t of cells of age τ , multiplied with the number ρ (t ,τ ) of cells of age
τ , divided with the mass mcb (t ,τ ) of the newly born cells from cells of age τ , integrated over
11
all cell ages τ , i.e. from zero to infinity.
A more simple relation than (2.14) will be used in this article. By assuming that all cells are born
with the same mass mc0 at all times, it follows from (2.12) and (2.14) that
mod ∞
mod
n
0
b

ρ (t , 0) = ∫ c1mc (t ,τ ) − c2 mc (t ,τ ) g (τ ) ρ (t ,τ ) / mc dτ , mc (t ,τ ) = mc0


ε
(
)
(2.15)
The equation set is now closed. A stepwise algorithm is stated in appendix A.
3 Analytical solutions
This section presents some analytical solutions when the cells starts to divide immediately at
time zero, the death coefficient µ (t ,τ ) is constant through time and independent of age, and the
mass mcb (t , τ ) of a newly born average cell is constant through time and independent of age, i.e.
µ (t ,τ ) = µ , mcb (t ,τ ) = mc0 , g (τ ) = 1,α = 1, β = 1, χ = 1.
(3.1)
Using (2.10) and (2.12), integrating (2.7) with respect to τ gives
∞
∞
N T (t ) = ∫ r (t ,τ )dτ − µ NT (t ) = (1/ mc0 ) ∫ (c1 − c2 )mc (t ,τ ) ρ (t ,τ )dτ − µ N (t )
0
0
= (c1 − c2 ) NT (t ) − µ NT (t ),
(3.2)
which has the exponential growth solution
NT (t ) = NT (t0 ) Exp[(c1 − c2 − µ )(t − t0 )],
(3.3)
where a dot above a variable means time derivation. NTd (t ) is the number of dead cells, and
NTad (t ) is the total number of cells (dead or alive), i.e. NTad (t ) = NT (t ) + NTd (t ) . The following
solutions follows directly from (3.3)
12
N Td (t ) = µ NT (t ) ⇒ NTd (t ) =
t
µ
∫ µ NT (t ')dt ' = c1 − c2 − µ [ NT (t ) − NT (t0 )],
t0
NTad (t ) = NT (t ) + NTd (t ) =
(c1 − c2 ) NT (t ) − µ NT (t0 )
c1 − c2 − µ
(3.4)
The fraction of dead cells of the total number of cells is then given as
NT (t )
NTad (t )
=
µ ( NT (t ) − NT (t0 ))
N T (t )
µ
=
, limt →∞ ad
, when c1 − c2 − µ > 0
(c1 − c2 ) NT (t ) − µ NT (t0 )
NT (t ) c1 − c2
(3.5)
Observe from the solution in (3.5) that increasing the death coefficient µ increases the fraction
of dead cells µ /(c1 − c2 ) and vice versa. This does not mean that the proliferation of cells is
stopped, since to stop the proliferation one must achieve that (c1 − c2 − µ ) < 0 as equation (3.3)
shows.
The analytical solution of the total number NTad (t ) of cells will be compared with simulation
results using the more general model in section 2, and with experimental results, in the next
section. The solutions given in (3.3)-(3.5) are typical during standard cancer treatments, where
use of chemotherapy and/or radiotherapy simply means to increase the death coefficient µ by
some fraction. We will show that the analytical solution can not in general be used to describe
the number of cells. This means that using digitoxin does not correspond to a simple increase of
the death coefficient µ .
4 Simulations
This section illustrates typical simulations of the model compared with in vitro experiments
with cancer cells exposed to digitoxin (Haux 1999). We show that the model matches the
experiments very well.
For the cell type we set mc0 to be a constant, which means that all cells at all times t are born
with the same mass. We further assume χ = 1 , which means that the mass growth of a cell
goes entirely into mitosis. We furthermore set that the death coefficient µ and the growth
13
parameter c1 are constants through time, i.e. that there is no change in temperature, oxygen
supply, and nutrino supply during a specific test. More specifically, without loss of generality
we set c2=0, since we can always redefine c1-c2 when n=1 in (2.12).
We first simulate the case without digitoxin. The numerical value of the parameters are found
by estimating values of the parameters c1 and µ such that the total number of cells (dead or
alive) are in agreement with the measurements. Thereafter we make the crucial assumption
that introducing digitoxin at subsequently higher concentration levels only changes the
numerical values of the parameters. For each subsequently higher digitoxin concentration we
estimate new values for the parameters, but change the values for as few as possible. The
number of cells (dead or alive) are again compared with the experiments.
Number
of cells
Exp T47D
6
1× 10
800000
600000
400000
Ana .
1
2
3
4
5
6
Time @ days D
Fig. 1: Analytical and experiments results for the number of cells (dead or alive) without
digitoxin, β = 0 / day, χ = 1 ,c1=0.6/day,µ=0.3/day,
mc0
= 4.2 10
−12
∞
kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
Fig. 1 shows the experimental and simulated results without digotoxin. The analytical and
experimental results correspond. The very good agreement between the analytical model and
the experimental results strongly simplifies the analysis, implying that the more general model in
section 2 is not necessary to invoke so far.
We now make the crucial step of introducing digotoxin such that the concentration is 25ng/ml.
digitoxin. Searching for values of the parameters which facilitate correspondence between
empirics and simulations for the number of cells (dead or alive), suggest that the analytical
solution given in section 3 does not fit the data. Fig. 2 and Fig. 3 show two attempts to fit the
data, but neither of them are good. But by using the general model in section with β = 1.3 / day ,
while keeping the other parameters constant as for the case without digitoxin, gives much
14
better fit to the data, as shown in Fig. 4. This is an important tentative observation; The
introduction of digitoxin gives an increased delay between the time of birth of a cell to the
time of mitosis. The values of the other parameters do not change!
Number of cells
700000
600000
500000
400000
300000
200000
1
2
Exp T47D
Ana .
3
4
5
6
Time @ days D
Fig.2: Analytical and experimental results for the number of cells (dead or alive) using 25 ng/ml
digitoxin, β = 0 / day, χ = 1 ,c1=0.6/day, µ=0.57/day.
∞
mc0 = 4.2 10 −12kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
Number
of cells
Exp T47D
1× 106
800000
600000
400000
Ana .
1
2
3
4
5
6
Time @ days D
Fig.3: Analytical and experimental results for the number of cells (dead or alive) using 25
ng/ml digitoxin, β = 0 / day, χ = 1 , c1=0.6/day, µ=0.3/day,
mc0
= 4.2 10
−12
∞
kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
Number of cells
700000
600000
500000
400000
300000
200000
1
2
Exp T47D
Sim .
3
4
5
6
Time @ days D
Fig.4: Simulation and experimental results for the number of cells (dead or alive) using
15
25ng/ml digitoxin, β = 1.3 / day, χ = 1 , c1=0.6/day, µ=0.3/day,
mc0
= 4.2 10
−12
∞
kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
Fig. 5-7 show the number of cells when the digitoxin concentration is 50ng/ml. Figs. 5 and 6
show two attempts to fit the data with the analytical solution. The fits are not so good. Fig. 7
shows the simulation results when the delay from birth to mitosis is increased to β = 5 / day ,
while the other parameters have the same values as in the case without digitoxin. Observe
now the very good agreement with the experimental results.
Number
of cells
350000
300000
250000
200000
150000
Exp T47D
Ana .
1
2
3
4
5
6
Time @ days D
Fig.5:Analytical and experimental results for the number of cells (dead or alive) using 50 ng/ml
digitoxin, β = 0 / day, χ = 1 ,c1=0.6/day,µ=0.7/day,
mc0
= 4.2 10
−12
∞
kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
Number
of cells
Exp T47D
1× 106
800000
600000
400000
Ana .
1
2
3
4
5
6
Time @ days D
Fig.6:Analytical and experimental results for the number of cells (dead or alive) using 50 ng/ml
digitoxin, β = 0 / day, χ = 1 ,c1=0.6/day,µ=0.3/day,
mc0
= 4.2 10
−12
∞
kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
16
Number of cells
350000
300000
250000
200000
150000
1
2
Exp T47D
Sim .
3
4
5
6
Time @ days D
Fig. 7: Simulation and experiments results for the number of cells (dead or alive) using 50
ng/ml digitoxin, β = 5 / day ,c1=0.6/day, µ=0.3/day,
mc0
= 4.2 10
−12
∞
kg , NT (0) = ∫ ρc0 dτ ≈ ρc0 ∆τ = 105 / ml
0
The tentative hypothesis based on the results for 25ng/ml digitoxin is strongly supported, i.e.
the introduction of digitoxin only enhance the time delay from birth to mitosis of the cells,
and higher values of digitoxin leads to larger time delays.
Fig. 8 assumes no exposure to digitoxin for the first 29 days, with a concentration of 50ng/ml
digitoxin at day 30 and for each day thereafter. Observe how the number of living cells starts
to flatten out at day 30 and thereafter.
Sim Number
of cells
Cells
1.2× 1099
1× 108
8× 108
6× 108
4× 108
2× 10
Living
10
20
30
40
cells
Time @ days D
Fig. 8: The number of cells when the digitoxin level is 50ng/ml after day 30.
A closer inspection shows that the critical concentration level is around 50 ng/ml digitoxin
for this cell line, which is not very sensitive to digitoxin ( Haux (1999)).
4 Conclusion
The article provides a mathematical description based on the theory of differential equations,
for the proliferation of malignant cells (cancer). A model is developed which enables us to
17
describe and predict the dynamics of cell proliferation much better than by using ordinary
curve fitting procedures. Theoretical results for the total number of cells (living or dead) is
found to be in good agreement with experiments for the cell line considered, assuming
different concentrations of digitoxin. The numerical results show that by introducing amounts
of digotoxin, the time from birth to mitosis increases, which effectively gives a mitosis delay.
The delay is probably connected with changes in the Ca ions concentration inside the cell.
The enhanced time between the birth and mitosis of a cell leads effectively to smaller
proliferation rates. This mechanism is very different from the mechanism appealed to by
standard chemotherapy and radiotherapy where the death ratios of the cells are mainly
affected. Based on the literature and the present results we have established the following
different mechanisms effectively reducing proliferation: a) necrotic death, b) apoptosis and c)
enhanced time between birth and mitosis. The last mechanism is as far as we know new.
By systematically analyzing and building models for the mechanisms appealed to by standard
treatment, and by use of digitoxin or other drugs which are likely to emerge, we expect that a
more specific treatment can be found for a given cell line, which increases the probability of
a successful treatment. The critical digitoxin level concentration, i.e. the concentration level
where the number of living cells is not increasing, is approximately 50 ng/ml for the cell line
we investigated in this article. It is most likely closer to 25 ng/ml for other malignant cell
lines (prostate cells for example; Haux 2000). Therapeutic plasma concentration of digitoxin
when treating cardiac congestion is about 15-33ng/ml, but individual tolerances are large.
The effect of digitoxin during cancer treatment is therefore very promising.
Aknowledgement
We thank dr. med P.K. Opstad and senior scientist S. Sterri at FFI for stimulating discussions
during the preparation of this manuscript.
References
Stenkvist, B., Bengtsson, E., Erikson, O., Holmquist, J., Nordin, B., Westman-Naeser, S.
(1979), “Cardiac glyco-sides and breast cancer”, Lancet 1, 563.
Bartilanffy, L. von (1955), “General Systems Theory,” Main Currents of Modern Thought I,
75-83.
Bartilanffy, L. von (1962), “General Systems Theory – A Critical Review,” General Systems,
18
Yearbook of the Society for General Systems Research 7, 1-20.
Haux, J., Lam, M., Marthinsen, A.B.L., Strickert, T., and Lundgren, S. (1999), “Digitoxin, in
non toxic concentrations, induces apoptotic cell death in Jurkat T cells in vitro”
,Originalia Z. Onkol./J. of Oncol. 31, 1, 14-20.
Haux, J., Solheim, O., Isaksen, T., Angelsen, A. (2000), “Digitoxin, in non-toxic
concentartions, inhibits proliferation and induces cell detah in prostata cancer cell
lines”, J. of Oncol. 32,11-16
Moxnes, J.F. and Hausken, K. (2003), “Predicting the Concentration Level of Anti-Cancer
Drug during Treatment of a Living Organism,” Medical Hypothesis 60, 4, 498-500.
Grossman ,M., Jamieson, M.J., Kirch, W. (1998),”Effects of digoxin and digitoxin on
circadian blood pressure profile in healthy volunteers”, Eur J clin Invest 28 701-6.
Russo, M.A., Isabella, M., Dani, A.M., Wolf, F. Cittadini, A. (1982),”Role of Ca++ and
Cytoskeleton
in control of cell shape”, Membranes in Tumour Growth, T Galeotti
et al, editors
Appendix A
Assume that all quantities are given at time t. The following algorithm is used for calculating
the values at time t + ∆t and τ + ∆τ ,where ∆τ = ∆t
∀τ ≥ 0,
ρ (t + ∆t ,τ + ∆τ ) = ρ (t ,τ ) − ∆t µ ρ (t ,τ )
ρ (0,τ ) = 0 when τ ≠ 0, ρ (0, 0) = NT (0) / ∆τ
(A1)
∀τ ≥ 0,
If
c1mc (t ,τ ) n − c2 mc (t ,τ ) ≤ 0, or
NT (t ) ≥ N max ,
mc (t + ∆t ,τ + ∆τ ) = mc (t ,τ ) + ∆t [c1mc (t ,τ ) n − c2 mc (t ,τ )]
else
mc (t + ∆t ,τ + ∆τ ) = mc (t ,τ ) + ∆t [c1mc (t ,τ ) n − c2 mc (t ,τ )][1 − χ (1 − Exp[−τ 2 / β 2 ])],
mc (0,τ ) = 0 when τ ≠ 0, mc (0, 0) = mc0
The boundary condition is given by
(A2)
19
∀t ≠ 0, ∆t ≠ 0,
If
c1mc (t ,τ ) n − c2 mc (t ,τ ) ≤ 0, or
NT (t ) ≥ N max , then
ρ (t + ∆t , 0) = 0, mc (t + ∆t , 0) = 0,
(A3)
else
ρ (t + ∆t , 0) = ∫
∞
0'
ρ (t ,τ ) (c1mc (t ,τ ) n − c2 mc (t ,τ )) χ (1 − Exp[−τ 2 / β 2 ])dτ / mc0 ,
mc (t + ∆t , 0) = mc0
The total numbers of living cells are
NT (t ) = ∫
∞
0
NTad (t )
∞
ρ (t ,τ )dτ , NTd (t ) = ∫ µ ρ (t ,τ )dτ dt ,
= NT (t ) +
0
NTd (t )
(A4)